cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
797
Views
5
Helpful
2
Replies

Nexus FEX extender

aamercado
Level 4
Level 4

Trying to wrap my head around the Nexus Data Centre solution compared to other vendors. Anyone make any comparison with other vendors. Anyways, FEX cons below and wonder if it is true or am I misunderstand the CCO doc.

1. FEX does not support local switching which seems inferior

http://www.cisco.com/en/US/docs/switches/datacenter/nexus2000/sw/configuration/guide/rel_4_0_1a/FEX-overview.html#wp1197054

2. Pinning seems weird concept:

http://www.cisco.com/en/US/docs/switches/datacenter/nexus2000/sw/configuration/guide/rel_4_0_1a/FEX-overview.html#wp1203877

so assume all 4 x 10Gig is used and per above link:

“When the Fabric Extender is brought up, its host interfaces are distributed equally among the available fabric interfaces. As a result, the bandwidth that is dedicated to each end host towards the parent switch is never changed by the switch but instead is always specified by you”

Note If a fabric interface fails, all its associated host interfaces are brought down and remain down until the fabric interface is restored.

Say 4/48 = 12 so for sake of argument, I assume

hostport 1-12 = Fabric/F1

hostport 13-24 = Fabric/F2 etc..

If Fabric/F1 dies, does all ports 1-12 die with it or does it get route over to F2, F3, and F4?

2 Replies 2

iyde
Level 4
Level 4

1. The FEX (or Nexus 2000) is equivalent to a line card in a Cat4500 as it does not have any processing power. All switching and configuring is made on the Nexus 5000 to which it is connected.

The FEX is meant to be a way of having a lot of 1GE ports for the Top-of-Rack but still as a part of a 10GE switch. Also in order to facilitate migration to a 10GE server environment.

2. you quote it yourself from the docu:

If a fabric interface fails, all its associated host interfaces are brought down and remain down until the fabric interface is restored.

So yes, the ports 1-12 die if your 10GE uplink die when you use pinning.

You can use EtherChannel as your uplink instead of pinning, as also described in the documentation. Then the load from the servers will be shared by the uplinks as in standard EtherChannel fashion.

Reason for using pinning? Could be that you would make sure that you know which uplink carries the traffic from which 1GE port.

HTH

Very helpfull :)

We have ordered 2x Nexus 5K's and 2 x Nexus 2 K's for a datacentre that we are building in July.

As this technolgy is so cutting edge it is beyond my cpmfort zone and we are bringing in a third party to configure to best practices.

Do you know of any training materials for the new unified fabric solutions as you clearly seem to know your stuff! This is something I would like to know more about as I have the perfect opportunity to learn.

Thanks.

Carlton

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Review Cisco Networking products for a $25 gift card